基于遗传算法改进BP网络的地表沉陷预计

SURFACE SUBSIDENCE PREDICTION IN MINED-OUT BASED ON GENETIC-ALGORITHM-OPTIMIZED BP NETWORK

本文采用遗传学习算法和误差反向传播算法相结合的混合算法来训练前馈人工神经网络,即先用遗传学习算法进行全局训练,再用BP算法进行精确训练,使网络收敛速度加快和避免局部极小。作为实例,将该方法应用于地表沉陷预计问题中。建立了采动地表沉陷的神经网络预计模型,利用矿区大量的地表沉陷实际观测数据样本对该神经网络进行训练和学习,并用该网络对几组数据进行采动地表沉陷预计。结果表明,该神经网络预计模型具有收敛速度快、预测精度高的优点,为采动地表沉陷预计提供了实用的方法。

A new method for training the artificial neural network is presented. In this method, the genetic algorithm （GA）, a general-purpose global search algorithm is used to train the network with updating the weights to minimize the error between the network output and the desired output. Then the back-propagation （BP） algorithm is used to further train the artificial neural network. The method is used to speed up the convergence and improve the performance. As an example, the method was used to predict surface subsidence in mined-out. An neural network prediction model was established. A lot of practical surveying data sets are collected to train the neural network. Several sets are used to predict the surface subsidence by the neural network. Results show that the neural network prediction model is of high convergent speed and good prediction precision, so the model offers a useful approach for surface subsidence prediction in mined-out.